MELD-Na and MELD Perform Similarly in Predicting Outcomes Post-TIPS

Abstract

Purpose: The Model for End-Stage Liver Disease (MELD) is a commonly used risk score to predict mortality in patients undergoing TIPS. Single center retrospective studies have found that, in addition to the MELD parameters, serum sodium (Na) is an important additional predictor of mortality. The aim of this study was to assess if the MELD-Na had a better mortality predictive value in patients undergoing TIPS. Methods: Consecutive patients undergoing TIPS between January of 2006 and December of 2010 were included. Medical records were reviewed and levels of sodium, creatinine, bilirubin, INR and albumin were abstracted. Outcome was death (obtained from the Social Security Death Index) or liver transplantation. The differences between groups were evaluated by medians of an independent sample t test or the Fisher exact test. Sensitivity, specificity, positive predictive value (PPV) and negative predicitve value (NPV) were calculated for the two models to predict 6 month mortality using two different cutoffs. The Somers D statistical test was used to compare the predictive power of both models. Results: Fifty patients were included. Twenty two patients (44%) died within 6 months after TIPS placement, and two (4%) underwent liver transplantation. The median MELD score before TIPS was 15 (7-41) among patients who died or received transplant vs. 9 (1-23) among survivors (p= 0.0044) and the median MELD-Na was 18.5 (11-41) vs. 12.5 (6-29)(p=0.0028), respectively. Sensitivity, specificity, PPV and NPV were calculated for the two models to predict 6 month mortality using two different cutoffs. With a score of 18 or above, MELD showed a sensitivity of 20%, specificity of 92%, a PPV of 71% and a NPV of 56% compared to MELD-Na that showed a sensitivity of 54%, specificity of 85%, PPV 54% and a NPV 85%. When the cutoff of 15 was used MELD showed a sensitivity of 54%, specificity of 85%, a PPV 54% and a NPV 85% compared to a sensitivity of 75%, specificity of 65%, PPV of 67% and NPV of 74% on the MELD-Na Score. There was no statistical significance with respect to gender, age, underlying liver disease and independent variables such as sodium, creatinine, bilirubin, INR and albumin between patients who survived and those dying or requiring liver transplant. The Somers D value for the MELD score was 0.537 (95%CI: 0.21-0.75, p=0.002) and 0.572 (95%CI:0.25-0.79, p=0.001) for the MELD-Na. The difference between models was not statistically significant. No early death was attributed to a fatal complication due to TIPS. Conclusion: In our population, MELD-Na performed similarly to MELD when assessing prognosis after TIPS placement and did not confer any significant advantage in identifying patients at high risk for 6-month mortality.Table: Comparing MELD to MELD-Na